This proposal responds to the research need to develop powerful biomarkers of disease response and the need to better understand current modalities of tuberculosis treatment. Multidrug-resistant tuberculosis (MDR-TB) has been declared a global emergency. Treatment outcomes are poor, driven by toxicity and limited efficacy of the 2nd-line antituberculosis drugs. Although there is evidence that both antituberculosis activity and most of the toxicity of the key drugs are related to drug exposure, the pharmacokinetic/pharmacodynamic (PK/PD) relationships in patients with MDR-TB are poorly characterized. In a cohort of South African patients with MDR-TB, we plan to: 1) use innovative pharmacometric analyses to develop a biomarker model of disease regression based on time-to-positivity in liquid culture of serial sputum samples; 2) describe the population pharmacokinetics of the five drugs constituting the standard treatment regimen in nonlinear mixed effects models; 3) describe the individual susceptibility of Mycobacterium tuberculosis isolates to those drugs, and the distribution to of the minimum inhibitory drug concentrations in the population; 4) define the frequency of drug-related side effects; 5) use computational analyses to quantify the individual contributions to efficacy and toxicity of the antituberculosis drugs used in combination. We propose to use in vitro hollow fiber models of tuberculosis (HFM-TB) to determine pharmacokinetic targets for antituberculosis activity and the suppression of resistance. Thus the (PK/PD) parameters associated with optimal efficacy, prevention of ADR and synergy will be investigated for moxifloxacin, pyrazinamide and kanamycin combinations in log-phase growth, semi-dormant and intracellular Mycobacterium tuberculosis, and then translated to patients using Monte Carlo simulations. The key PK/PD relationships defined in the clinical study will also be investigated in HFM-TB. Our study will identify doses and drug combinations for the treatment of MDR-TB that are likely to be more efficacious and less toxic than the currently used regimen. Moreover the development of a biomarker model of disease response and computational analytical methodologies will enable more efficient optimization of future regimens.